Olist Payments Analytics
Layered dbt analysis over ~100k Brazilian e-commerce orders to understand payment methods and installments.
PythonpandasBigQuerydbt Core
Architecture
Problem
Understanding how Brazilian consumers pay in e-commerce — credit card, boleto, vouchers and installments — matters for fintechs and benefits companies designing financial products.
Solution
ELT pipeline over the public Olist/Kaggle dataset (2016–2018, ~100k orders). Extraction with Python, load into BigQuery, and layered dbt modeling: 5 staging models, 2 intermediate and 3 analytical marts, with a documented DAG and full lineage.
Results
- Credit card accounts for more than 70% of transacted value
- Average credit ticket of R$ 163.32 over 3.51 installments, vs. R$ 142–145 upfront on boleto/debit
- 26 automated tests passing, with a documented DAG and lineage